Urban and nomadic isotopic niches reveal dietary connectivities along Central Asia’s Silk Roads

The ancient ‘Silk Roads’ formed a vast network of trade and exchange that facilitated the movement of commodities and agricultural products across medieval Central Asia via settled urban communities and mobile pastoralists. Considering food consumption patterns as an expression of socio-economic interaction, we analyse human remains for carbon and nitrogen isotopes in order to establish dietary intake, then model isotopic niches to characterize dietary diversity and infer connectivity among communities of urbanites and nomadic pastoralists. The combination of low isotopic variation visible within urban groups with isotopic distinction between urban communities irrespective of local environmental conditions strongly suggests localized food production systems provided primary subsistence rather than agricultural goods exchanged along trade routes. Nomadic communities, in contrast, experienced higher dietary diversity reflecting engagements with a wide assortment of foodstuffs typical for mobile communities. These data indicate tightly bound social connectivity in urban centres pointedly funnelled local food products and homogenized dietary intake within settled communities, whereas open and opportunistic systems of food production and circulation were possible through more mobile lifeways.


Supplementary Information 2: Subsistence review for medieval Central Asia
The food production systems of medieval Central Asia were supported by sophisticated irrigation technology as well as extensive rainfed farming and livestock management that, together, generated a diverse and plentiful supply of crops and animal products to both urban and nomadic realms [11][12][13][14][15] . Ancient texts of Chinese and Arab geographers describe lush Central Asian oases where wheat, barley, millet, rice, peas, chickpeas, lentils, melons, and cotton were cultivated, in addition to gardens, vineyards, and orchards of fruit and nut trees 1,16,17 . Incidental recovery of carbonized botanical remains from archaeological sites and ubiquity of agricultural artefacts confirm these cultigens were grown and processed throughout the region [18][19][20][21][22][23][24] . More systematic recovery of botanical remains confirms a rich diversity of cultivated taxa with a strong emphasis on various grains as staple foods [25][26][27][28][29] , alongside local pastoral production and, presumably, animal products acquired from local nomadic herders 18,30,31 .
The historical record further documents bustling fairs and markets in which foods were traded in bulk and also exchanged in quantities for individual households or estates. The geographic extent of this commerce has been documented through ceramics artefacts, which derive from major urban production centers in Samarkand, Tashkent, Bukhara, Afghanistan, Iran, China and beyond 1,6,18,31 . One of the most prominent markets took place in the Zaamin Mountains., in a city called Marsmand, which was supposedly located next to a major centre for mining ore and metal production named Mink 18,19 . While existing textual accounts and subsistence datasets have provided insights into interactions with food in urban centers and fortified settlements, no published data exists to elaborate upon subsistence for nomadic pastoralists in later medieval times. Inconsistent chronologies, geographic coverage, and recording methods preclude using existing data to understand regional dietary diversity and food exchange tied to the historical record.

Supplementary Information 3: Sampled sites, archaeological context, and subsistence data Uzbekistan
Khoresm Tok-kala is located near the modern city Nukus in Khoresm. The fortress, settlement, and cemeteries that comprise the preserved site cover approximately eight hectares and are located on a raised hill above the floodplain of the Amu Darya. An ancient river channel runs along the base of the hill, and there is evidence of multiple flooding events throughout its history, which appear to have destroyed much of the settlement architecture. Flooding in Khoresm is periodic, which has catastrophic potential to alter the course of the Amu-Darya and affect the level of the Aral Sea, and the harnessing of these floodwaters is considered to be a critical component to the success of irrigation systems in Khoresm 10 . During periods of vast hydrological change, such as in the 4 th c., settlement patterns were drastically altered and sites were abandoned 32 .
Archaeological materials recovered at Tok-kala seem to indicate significant interaction with nomadic groups in the region 21 . Besides a modern Muslim cemetery, there are three known cemeteries at Tok-kala 21 . The first is an early medieval (7 th -8 th c.) naos with ossuary burials, consistent with Zoroastrian practices. Notably, this phase contains well-preserved pieces of wall paintings and inscriptions on ossuaries that have led to speculation about their cultural origins, either directly from Sassanid roots or from older heritages pre-dating the Sassanid period 33 . The other two burial grounds are middle medieval (9 th -11 th c.) Muslim cemeteries, identified from grave forms and body position of interred individuals. Samples in this study were recovered from individuals in Muslim burials. The first of these cemeteries is located on top of the earlier Zoroastrian burials. Despite its relatively small size, Tok-kala is sometimes known as a 'besh-kala,' or mother-city, of Khoresm, indicating past cultural significance. Tok-kala is considered to have been a regional centre of political and economic influence 21 .
The most comprehensive information about Tok-kala can be found in Gudovka's monograph 21 . She reports the recovery of wheat, barley, and millets from multiple cultural layers, but these macrobotanical remains were not quantified. Rich livestock and fish bones were also reported but were not quantified.

Tashkent Oasis
Uturlik sits on the left bank of the Syr Darya River in the Sirdarya Region of modern Uzbekistan, near Tashkent. In the historical and archaeological literature, the ancient city and local region are often referred to as Chach 31,34 . The site area covers approximately 60 hectares, and was occupied from the 6 th -12 th c. 35 , and our samples from the site date to the 10 th -12 th c. A large medieval Muslim cemetery of approximately 4,000 graves is located on the western side of the site. The large area of the settlement and cemetery at Uturlik have led researchers to estimate a total population of between 8 and 10 thousand people at its peak 35 . The city is described as having a sprawling residential and artisan sections that provided a diverse and productive craft economy, in addition to vibrant markets 31 . Despite the apparent size of Uturlik, there is a paucity of detailed information in published sources about the ancient city. Scholars highlight the location of the city as being on an important trade route between Samarkand, to the south, and the Otrar Oasis, located in Kazakhstan, which has led to speculation about interactions between settled and nomadic populations 31 . There does not appear to be any quantified subsistence information available for Uturlik, but scholars describe the ancient city as being supplied by irrigated and rainfed fields, in addition to livestock herding in the local hilly landscape 31 . There is the possibility that fishing occurred in the marshes, irrigation canals, and Syr-Darya river 31 , but there is no material evidence for fishing reported in the literature.

Ferghana Valley
Chor Dona is located in the southeastern region of the Ferghana Valley, on the northwestern outskirts of the city of Andijan. The site was occupied during two periods: 7 th -8 th c. and 14 th -16 th c. 36 . Between these two occupations, the site was used as a cemetery. The majority of burials at Chor Dona are typical of Muslim practice, but there are at least five individuals buried with objects and grave structures similar to pre-Islamic rituals common in medieval sites in Kazakhstan associated with nomadic groups 36 . Chor Dona's proximity to ancient Andijan would have likely put its residents within the economic and social sphere of this large urban centre.
Chor Dona appears to have served as a large-scale economic production centre for grains, as evidenced by flour milling facilities 36 . Macrobotanical remains of foods recovered from the site include walnuts, millet, barley, and wheat, in addition to an assortment of agricultural tools associated with cultivation, harvesting, and processing 36 . Furthermore, fruit seeds of grapes, cherry plums, peaches, and apricots were reported as ubiquitous in the site 36 . Cotton seeds were also recovered 36 .
Chartok is not well described in the archaeological literature. The site was published in a craniometric study on medieval populations in Uzbekistan that briefly describes Chartok as dating to the 12 th c. and located in the Namangan region of northern Ferghana 34 . Unfortunately, no cultural context is given about the site. Searches for additional information did not yield results.

West Pamir-Alay
Tashbulak is a recently discovered site located in the highlands of the Malguzar Mountains, close to the border with Tajikistan, at approximately 2100 meters above sea level 37 . It is the only known highland centre that was constructed and occupied during the Qarakhanid Empire. This region has historically been occupied by nomadic groups, but during the 9 th -13 th c., a series of small cities and large villages were constructed in the mountain zone 38 . Tashbulak consists of an elevated citadel above a lower town area with workshops and possible residencies 9 . The site also contains a large Muslim cemetery of approximately 400 burials 9 . The site spans from the 9 th to early 12 th c., which is currently dated according to coins and Qarakhanid ceramics 9 . Research at the site is ongoing, but the presence of iron objects, glazed and wheel made ceramics, and non-local fruits indicate the likelihood that Tashbulak served as an important market, trading post, and/or production centre between highland and lowland regions. Macrobotanical remains of grains recovered from the site include barley and wheat, and millets were not found 9 ; detailed paleobotanical work is forthcoming by Robert Spengler.
Altyntepe is in the western part of the Kashgardarya Oasis, near the modern-day town of Karabag. Although it was founded in the 7 th -8 th c., the site reached its peak in the 10 th -13 th c. and covered about 40 hectares 39 . The city consisted of a fortified inner city with surrounding sprawling settlements, and a cemetery is located on the southwest border of the site 39 . The remains of several different craft workshops have been recovered, including industrial-scale brick and pottery production complexes 39 . Despite evidence of economic fluorescence, some researchers point to the relatively small defensive walls as an indication that Altyntepe was not as important or wealthy as other medieval cities, such as Penjikent 39 .
Macrobotanical remains at Alyntepe include watermelon, muskmelon, and grape seeds 39 . Excavations recovered bulk remains of wheat, which were reported as club wheat (Triticum compactum) 39 . Barley was not reported from Altyntepe, but small quantities were found in nearby sites 39 . Animal bones present in ash deposits were found throughout the site 39 . While fish bones were not reported, numerous ceramic bowls and dishes were recovered that were decorated with stamp impressions depicting slender symbols of fish 39 .
Frinkent, also known as Afarinkend, is located in the lowlands of the Zerafshan valley, in the interfluve between the Kara-Darya and Ak-Darya rivers. Frinkent is sixty kilometres northwest of the city of Samarkand. The site, delineated by a wall, includes a fortress complex, cemetery, and settlement area covering approximately 14 hectares 40 . The site was founded in the 7 th century by the brother of the ruler of Samarkand, making it a site of potential political importance 40 . The site continued to be an important centre of the Zoroastrian faith in the Zerafshan region even through the 10 th century, when Islam had replaced most other faiths 41 . The cemetery at Frinkent is a stark example of this role. Hundreds of individuals were cremated or interred in large ceramic jars, some of which were painted on the exterior and had soot deposits as would have been caused by burning 40 . Many of the jars contained the skeletal remains of multiple individuals, and often the mandibles were missing, indicating skeletonization before primary or secondary deposition in the vessels 40 . Subsistence data could not be located for Frinkent.

Otrar Oasis
Konyr-tobe, also known as Kuyuk-Mardan, is a medieval fortress located on the Arys River. Two necropolises remain the only components of the site that have been systematically researched and published 42 . Konyr-tobe I contains 120 burials that were placed on a unique platform mound constructed from local clay-rich soil and mudbricks, which stands about 2.5 m above the ground level 42 . The platforms is outside of the fortress, and its association to the fortress is unclear. The human remains with stable isotopic data included in our study are sourced from these burials 43 . Located about 40 m from Konyr-tobe II is a similar but smaller platform mound with 8 burials 42 . Regional specialists consider these platform cemeteries, which are widely distributed in the Arys valley and Otrar Oasis, to have been built in order to stand seasonal floods in the area that heavily saturate the takyr surface 44 . This style of cemetery construction is unique to this site and scholars have interpreted it to represent nomadic communities 42 .
Konyr-tobe I was in use from the 5 th -7 th c. based on iron weapons and implements, in addition to ceramic vessels, interred with the bodies 42,44 . Burial practices show resemblance to pre-Islamic and late medieval rituals, despite artefacts largely reflecting an early medieval chronology 42,44 . This burial variability suggests high levels of interaction with outlying communities, which is further demonstrated by distinctive ceramic vessels with spouts common in the Konyr-tobe assemblage that were also found nearly 800 km away in contemporaneous layers of a nomadic encampment site in the desert-steppe margin of the Dzhungar Mountains, named Mukri 45 . In rooms of a nearby city, Kok-mardan, ceramic storage vessels were found with millet, wheat, barley, rice, and peas 42 , but subsistence data for Konyr-tobe or its cemeteries were not reported.
Temirlanovka is an early medieval burial site located on the northern bank of Arys River. The site contains 49 burial mounds that date to the 2 nd -4 th c. based on material inventories of the graves 46 and one individual that was directly radiocarbon dated (cal AD 140-376) 43 . Most of the site remains unpublished, but an iron dagger from burial #33 has raised speculation that the community using the Temirlanovka burial ground had contacts with nomadic groups as far away as the Black Sea region, which during the late Iron Age can be generally described as a territory of the Scythian-Sarmatian cultural groups 47 . Pollen analysis from sediments recovered from ceramic vessels found in the burials revealed natural background vegetation of Ephedra and Artemisia, in addition to a small number of pollen grains from buckwheat and legumes 46 .

Zhetysu (Semirech'ye)
Turgen is a multi-period settlement and burial complex in the modern city Turgen. The site contains mostly late Andronovo to Wusun cultural phases, but there is are also early medieval burials, which further extend across the entire floodplain the Ili Valley 48 . The burial forms include low earthen mounds surrounded by stone fences, which are characteristic of the Turkic period 48 . The subsistence economy of the early medieval period is believed to be similar to that of the late Iron Age, which included dense villages, irrigation agriculture, and management of livestock 49 . However, detailed excavation data for the site is not well reported. The chronology of the medieval burials is estimated to span the 2 nd -6 th c., according to burial forms and material remains that indicate nomadic traditions.
Butakty is а multi-period settlement and burial complex in the modern city Almaty. The site includes occupations from the late Bronze Age Andronovo cultural horizon (ca. 1700-1200 BC) to the medieval period (ca. AD 12 th c.) 50,51 . The settlement areas contain rich artefact assemblages and architectural remains that indicate intensive, perhaps year-round occupation at the site 50,51 . The medieval burials are reported to be low in number, but they contain evidence of ritual practices common in the Zhetysu region, which includes burying sheep/goat limbs with the human bodies 51 . One individual from Butakty was directly radiocarbon dated to cal AD 1022-1155 43 . Iron fasteners and ring pendants share style with those characteristic in nomadic groups in central Kazakhstan and the Altai region 50,51 . No subsistence data could be located for Butakty, except brief mentions of an agro-pastoral economy that persisted throughout the site's occupational sequence 50,51 . (Butakty had been previously published with the incorrect geographical coordinates that placed it 10 km east of Turgen 43 , but its location has been corrected in our paper, as shown in main text Fig. 1 and Table 1

.)
Karatal is a multi-period, ancient cemetery complex located on the foothills of the Dzhungar Mountains and is about 1 km from the Karatal River. The site underwent partial excavation and survey in 2006 by Alexei Mar'yashev, Taylor Hermes, and Marc Meyer, and three periods of use were identified (unpublished report). In 2008, another campaign continued work on medieval burials. The first occupation occurred during the late Bronze Age (ca. 1700-1200 BC), as indicated by distinctive stone burial cists that contained cremated human remains and 'Andronovo'-style ceramics. Some of the burial cists were arranged in pairs and enclosed within a low-profile stone fence. The next occupation occurred during the Iron Age (ca. 1000 BC -AD 300), as indicated by linear rows of earthen mounds (kurgans) placed on the flat areas of rolling hills. These kurgans were not excavated. Within the site's environs, numerous rock outcrops were found to be decorated with panels of Bronze and Iron Age petroglyphs that depict animals and anthropomorphic figures, which are common in the Zhetysu region 52,53 . Abandoned settlements, consisting of stone foundations for small structures (likely felt tents), were recorded on survey at low density, which suggests local residence by pastoralist groups.
The final phase of the Karatal cemetery complex occurred during the medieval period. Four adjacent burials containing the remains of children (0-3 years) radiocarbon dated to the 17 th -19 th c. (unpublished report). These were not included in the study because of their recent chronology into the early modern period and young age. Juveniles exhibit enriched δ 13 C and δ 15 N values relative to adults, due to trophic effects from gestation and nursing 54,55 . Children also tend to eat different foods than adults 56 .
An individual recovered from an 8 th -10 th c. kurgan-style burial was included in this study. The low-lying stone mound was located seven meters from the cluster of early modern burials. A looted context was found that contained disarticulated remains of a female individual (KZ-BR5), which was directly dated to cal AD 773-916 (2-sigma; OS-72514; 1160 ± 30 BP), which is first reported here. The other two medieval individuals that were included in this study were excavated in 2006. Two individuals (Karatal BR-1 and BR-2) were excavated after discovering the remains were partially exposed and threatened by an advancing erosional cut in the topsoil and deeper strata. Karatal BR-2 was directly radiocarbon dated to cal AD 972-1032 57 , and Karatal BR-1 is presumed to be from the same period based on stratigraphic positioning and associated horse riding tack and iron weapons, similar to those found in burials in the medieval period 51 . Around one of the leg bones of the Karatal BR-2 individual, the excavators found a cooper band that had preserved two small pieces of woven cotton 57 .
The finding of cotton fabrics suggests that cotton may have been an exchange good along the Silk Roads during the 10 th -11 th c., as it is possible that cotton could not have been grown as far north as Zhetysu 57 . However, medieval agriculture in Zhetysu may have supported water demanding crops such as cotton, on account of rice being grown to this day in the Karatal River watershed 58 and rice grains were recovered from medieval sites in the Otrar Oasis 25 . The only subsistence data recovered from excavated Karatal burials included a small number of sheep bones in KZ-BR5 that represent one individual animal, and a horse burial next to Karatal BR-1 that was not fully excavated.

Dehistan Plain
Geoktchik Depe is an urban complex of approximately 5.5 hectares that comprises two main features: a large mound that contains much of the Iron Age occupation (13 th -8 th c. BC) and a rectangular structure that is associated with the Sassanid and Islamic period (6 th -16 th c.) 59 . The site contains a rich faunal assemblage, which has been published in detail, and suggests a strong pastoral component to subsistence 59 . Notably, a small number of fish and bird bones were recovered from several phases of the site 59 , which had been incorrectly reported in a paleodietary study presenting a stable isotope analysis 60 . The site is associated with a complex network of irrigation canals that likely supported intensive farming 59 , which appears to have been dominated by C 3 crops (wheat and barley) that primarily contributed to human diets at the site 60 .
Misrijan is not well described in the English archaeological literature beyond a mention of being a 'medieval town' 59 , and a search for specific information about the site was unsuccessful. Human and fauna carbon and nitrogen isotopic values are reported from the site, but cultural details are not provided 60 . There is a detailed monograph in Russian that describes the medieval archaeological sites on the Dehistan Plain 61 . However, the naming of the site under analysis as 'Misrijan' by Bocherens et al. 60 may be confused with the Misrijan Oasis, which contains dozens of large medieval settlements and precludes identifying the precise site at this time. Figure S1: Boxplots of carbon and nitrogen stable isotopic ratios from all regions and periods.

Supplementary Information 5: Environmental modelling
Environmental conditions at sites were estimated from abiotic variables using an approach combining geographic information system (GIS) and redundancy analysis (RDA). Both 10 and 50 km analytical buffers were created around each site, which was used to calculate two sets of spatial means of various climatic and soil parameters derived from geographic data layers (Supplemental Tables S5-S6). Supplemental Fig. S2 shows examples of annual mean temperature, precipitation of wettest and driest month, and weight % of clay in soil. Climatic data were obtained from the WorldClim project (http://worldclim.org/), which describe biologically meaningful variations of temperature and precipitation in ca. 1 km resolution (bioclim) 62 . These data represent monthly climate records from 1970-2000. This period is similar in average temperature and precipitation to the 'Medieval Warm Period' during the 8 th -13 th c. in Central Asia [63][64][65] . However, the early medieval period (2 nd -6 th c.), which is chronologically represented in our dataset, coincides with the 'Dark Ages Cold Period' and is less comparable to modern environmental patterns 66 . Soil data were obtained from the Soil Grid project (http://isric.org/) to provide additional variables relevant for agricultural production 67 . The size of the buffers serves three purposes: 1) represent two spatial scales of human life experiences, either through direct movement and interaction or through contact with others living in this zone; 2) capture environmental variation in site environs caused by landscape fluctuations, such as mountains acting as rain shields and having steep, rocky terrain that changes over short distances, which was especially evident for areas near Tashbulak, Butakty, and Turgen; 3) overcome differences in spatial scales of the underlying geographic data.
Environmental parameters for each site buffer where associated with that site's δ 13 C and δ 15 N mean, min, and max values. The mean isotope values, as point estimates, were insufficient to represent sites expressing wide isotopic variation, such as Konyr-Tobe in δ 13 C or Temirlanovka in δ 15 N; min and max values were included to consider values at the ends of wide isotopic distributions, which were used in RDA with the R package vegan 68 . In order to reduce the susceptibility of the RDA to false positives, statistical significance was set at p < 0.01. It is important to note that this figure is considered liberal by scientists from multiple fields, who recommend p < 0.005 69 . Environmental parameters were examined for correlation and plotted as a distance matrix using hierarchical clustering (Supplemental Fig. S3). Highly correlated parameters were removed, and the remaining parameters were accepted to be used in RDA:   10km  50km  elevation  elevation  bio2  bio1  bio3  bio2  bio4  bio3  bio6  bio4  bio9  bio15  bio12 CECSOL CLYPPT CLYPPT Using these parameters, stable isotopic data were fitted to each RDA using the envfit() function in vegan. No statistical significant relationships were found at p < 0.01 (Supplemental Fig. S4). However, for the 10km dataset at p < 0.05 a negative correlation was found between mean δ 13 C values and bio3 (isothermality), in addition to a positive correlation was found with bio4 (temperature seasonality). For the 50km dataset, a negative correlation was also found between accepted environmental parameters and bio3.
To further explore possible relationships, multiple linear regressions were performed using stable isotopic data and accepted environmental parameters. Critically, the Bonferroni correction was applied to the level of significance in order to counteract the problem of conducting multiple, simultaneous statistical tests, which would otherwise increase the probability of finding an erroneous result 70  Isothermality is a measure of the difference between day-to-night temperatures relative to the annual summer-to-winter temperatures. A value of 100 indicates that the diurnal temperature range is equal to the annual temperature range, and smaller values indicate lower temperature diurnal ranges against the annual range. C 4 plants, including millets used in agriculture, are well adapted to hot and arid climate 71 . Locations with lower isothermality values have higher human mean δ 13 C values, but it is not clear why isothermality would correlate with human δ 13 C values. The chances of this correlation being spurious is supported by a clustering of sites in the bi-plot in three groups along the trend line and an overall low range of isothermality values between ca. 27 and 35% (Supplemental Fig. S5), which suggests there is unlikely to be a dependent relationship with factors that would influence dietary intake of C 4 plants.
There also appears to be a weak, negative linear correlation between elevation and mean δ 15 N values, but with a p-value of 0.033 and ca. 32.5% data explanation (R 2 = 0.325), confidence in this result should be checked. In hot and arid ecosystems, typical of lowland regions in Central Asia, plant δ 15 N values increase due to nitrogen loss from soils 72 . Agricultural products in lowland zones, especially those downstream from further farming run-off, can also be enriched in 15 N from exogenous inputs such as fertilizers and animal waste [73][74][75][76][77] . Thus, it is not possible to separate these processes given the prevalence of highly modified agricultural landscapes in the medieval period of Central Asia. Furthermore, these results demonstrate that patterns of agricultural production and dietary intake are more likely to be minimally affected by inter-regional differences in environment that were modelled here. Supplementary Figure

Supplementary Information 7: Isotopic niche overlap of urban medieval regions in Uzbekistan and
Turkmenistan SIBER Bayesian ellipse overlap analysis depicted in main text Figure 3, (Dehistan included below):